基于UAV-LiDAR点云的地表三维形变同步提取方法研究  

Research on the method of synchronous extraction of 3D surface deformation based on UAV-LiDAR point cloud

作　　者：占喜林 周大伟[1] 安士凯 詹绍奇 刁鑫鹏[1] ZHAN Xilin;ZHOU Dawei;AN Shikai;ZHAN Shaoqi;DIAO Xinpeng(School of Environment and Spatial Informatics,China University of Mining and Technology,Xuzhou,Jiangsu 221116,China;Anhui Coal Mine Green and Low-carbon Development Engineering Research Center,Ping'an Coal Mining Engineering Technology Research Institute Co.,Ltd.,Huainan,Anhui 232033,China)

机构地区：[1]中国矿业大学环境与测绘学院,江苏徐州221116 [2]平安煤炭开采工程技术研究院有限责任公司安徽省煤矿绿色低碳发展工程研究中心,安徽淮南232033

出　　处：《中国矿业大学学报》2023年第6期1241-1250,共10页Journal of China University of Mining & Technology

基　　金：国家自然科学基金项目(52104174);江苏省自然科学基金项目(BK20190642);宁夏回族自治区重点项目(2022BEG03065)。

摘　　要：针对地表沉陷是时空同步的三维移动过程,而常规监测处理是将下沉和水平移动分开提取,该方法在带来误差的同时也无法反映变形的同步性.本文在无人机光探测与测距(UAV-LiDAR)点云的基础上,提出一种刚体点云变换的方法来同步提取地表三维移动变形,主要步骤为:对LiDAR点云进行连通域分割得到地面特征点云;以地面特征点云为刚体,联合快速点特征直方图(FPFH)和迭代最近点算法(ICP)对其配准并从变换矩阵中提取地表三维移动;采用多尺度模型到模型的点云比对(M3C2)算法改正配准误差,以鄂尔多斯王家塔煤矿为例进行应用并评定精度.试验结果表明:地表最大下沉为2.723 m,X和Y方向的最大水平移动量分别为0.65和0.75 m;地表三维形变内部符合精度为57 mm,而M3C2算法对形变误差改正后的精度为41 mm,精度提高了28%.对比实时动态载波相位差分(RTK-GPS)实测形变值,X,Y和Z方向的形变中误差分别为56,83和36 mm,验证了基于点云分割配准提取三维形变是可行的.The surface subsidence is a three-dimensional movement process synchronized in time and space,and the traditional monitoring process is to separate the subsidence and horizontal movement.This processing method brings errors and cannot reflect the synchronization of deformation.Based on the UAV-LiDAR point cloud,this paper proposes a rigid point cloud transformation method to simultaneously extract the three-dimensional movement deformation of the ground surface.The main steps include:connecting domain segmentation of LiDAR point cloud to obtain ground feature point cloud;The ground feature point cloud is regarded as rigid body,FPFH combined with ICP algorithm was used to register the rigid body point cloud,and 3D ground point movement was extracted according to the transformation matrix;The M3C2 algorithm was used to correct the registration error.Take a Wangjiata coal mine of Ordos as an example to apply and evaluate the accuracy.The results show that:the maximum surface subsidence caused by mining is 2.723 m,and the maximum horizontal movement in X and Y directions is 0.65 and 0.75 m,respectively.The internal coincidence accuracy of the three-dimensional surface deformation is 57 mm,while the deformation error corrected by the M3C2 algorithm is 41 mm,and the accuracy is increased by 28%.Compared with the measured deformation values of RTK-GPS,the deformation errors in X,Y and Z directions are 56,83 and 36 mm.Therefore,it is feasible to extract 3D deformation based on point cloud segmentation and registration.

关 键 词：UAV-LiDAR 刚体点云 连通域分割 配准 地表三维形变 

分 类 号：TN29[电子电信—物理电子学]